Researchers at MIT have developed a new attack called TONTOU that exploits a vulnerability in Intel and AMD CPUs, bypassing existing Spectre defenses. This attack leverages timer interrupts to reopen a window for branch predictor poisoning, a technique used to trick CPUs into executing unintended code. A proof-of-concept exploit targeting Zen 2 processors has been demonstrated, highlighting a significant security concern for modern processors. AI
IMPACT This research highlights a potential vulnerability in the hardware underlying AI systems, which could impact the security and reliability of AI computations.
RANK_REASON Security research paper detailing a new CPU attack. [lever_c_demoted from research: ic=2 ai=0.4]
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →